Feng Liu (he/him) -- Assistant Professor at The University of Melbourne


Feng Liu

Feng Liu (he/him), Ph.D.

Assistant Professor in Statistics (Data Science),
School of Mathematics and Statistics, The University of Melbourne

Visiting Scientist @ Imperfect Information Learning Team,
RIKEN Center for Advanced Intelligence Project (RIKEN-AIP)

Visting Fellow @ DeSI Lab,
Australian Artificial Intelligence Institute, UTS

Address: Room 108, Old Geology Building (South Wing),
Building #156, Monash Road, Parkville VIC 3052, Australia.
E-mail: fengliu.ml [at] gmail.com or feng.liu1 [at] unimelb.edu.au
Phone: +61 3 9035 3645
[Google Scholar] [Github] [Group Website]


    I am always looking for self-motivated PhD, MPhil, Research Assistants, and Visiting Researchers. Please see this page for recruiting information, and check this page for the school information. Meanwhile, I am happy to host remote research trainees. You can collaborate with many excellent researchers in the frontier machine learning research areas in our group. Check this page for more information.


Research Interests

    My research interests lie in statistical hypothesis testing and trustworthy machine learning. Specifically, my current research work center around the following topics:
    Statistical Hypothesis Testing:
  • Two-sample Testing: Testing if two datasets are drawn from the same distribution.

  • Goodness-of-fit Testing: Testing if data are drawn from a given distribution.

  • Independence Testing: Testing if two datasets are independent.

    Trustworthy Machine Learning:
  • Defending against Adversarial Attacks: Detecting adversarial attacks (i.e., adversarial attack detection); Training a robust model against future adversarial attacks (i.e., adversarial training).

  • Being Aware of Out-of-distribution Data: Detecting out-of-distribution data; Training a robust model in the open world (e.g., open-set learning, out-of-distribution generalization).

  • Learning/Inference under Distribution Shift (a.k.a., Transfer Learning): Leveraging the knowledge from domains with abundant labels (i.e., source domains)/pre-trained models (i.e., source models) to complete classification/clustering tasks in an unlabeled domain (i.e., target domain), where two domains are different but related.

  • Protecting Data Privacy: Training a model to ensure that the training data will not be obtained by inverting the model (i.e., defending against model-inversion attacks).

Research Experience


  • Ph.D. in Computer Science (November 2020)

  • Faculty of Engineering and Information Technology,
    University of Technology Sydney, Sydney, Australia.
    Supervised by Dist. Prof. Jie Lu and Prof. Guangquan Zhang

  • Master of Science (June 2015)

  • School of Mathematic and Statistics, Lanzhou University, Lanzhou, China
    Supervised by Prof. Jianzhou Wang

  • Bachelor of Science (June 2013)

  • School of Mathematic and Statistics, Lanzhou University, Lanzhou, China


Australian Research Council